Excellence · Innovation · Scale

I solve real problems, prove the solutions work, and scale them to millions.

I’m CPO of Metaintro, a job-search AI company in New York. I designed the engine that reads the open web into labor-market data, and the memory system that lets our AI agents work as a team. Before this I was Director of Engineering at Moody’s Analytics on Climate on Demand, its climate-risk platform, and a software engineer at JPMorgan Chase, working on commercial real estate systems. I also wrote Keep a Value, a published standard for stating what a piece of shipped work changed, and how to check it.

New York · the open web, conducted

The open web, conducted. This hero is an animated field — web chaos resolving into structured labor-market rows — and needs JavaScript. The story it tells is the headline above.

Now — building AI job search, and a memory system for every seeker

Selected work

all →

Metaintro · designed it

Web-intelligence engine

Reads the open web into structured labor-market data — built toward 500M+ domains.

Metaintro · created it

AI-team memory system

The memory, ticket & file layer a team of AI agents runs on.

in production daily

Moody’s · Director of Engineering

Climate on Demand

ESG & physical-climate risk analytics for real-estate portfolios.

institutional scale

Solo · creator

Milo

A personal AI assistant for my family — live across SMS, WhatsApp, voice, web, and push.

gomilo.app

How I work

Build

Start with a real problem someone actually has, and build the system that solves it end to end — the data, the product, the operations around it. Not a demo of the easy half. Most of the real work hides in the parts nobody shows.

Prove

Ship it, measure it, and verify it before calling it done. The verification has to come from outside the work: write the check before claiming the result, hand the thing to someone with no context and see if it stands, keep the unflattering numbers next to the good ones.

III

Scale

Only what survives proof gets to grow — more people, more data, more load, on the same system, with the checks still passing. Scaling something unproven doesn’t make it better; it makes the failure bigger.

Writing

all →

A library of documents meant to be useful on their own — the standards I work by, and records of what building this way actually looks like.

The stranger testHand your work to someone with zero context and let them reconstruct what it's for. The gaps are the defects.2026 Green lights that lieThe anatomy of verification signals that pass while the thing is broken, and how to build checks that can't.2026 My engineering team is a fleet of AI agentsMost days my engineering team is AI agents. What makes that work is not the models. It's measurement, enforcement, and refusing to take an agent's word for anything.2026 Job search is an information problemMost job-search pain gets billed as a motivation problem or a skills problem. It is mostly neither. It is an information asymmetry.2026

Ask me anything

Questions arrive privately; the ones worth answering get answered publicly.

Ask a question